Are the feds handing OpenAI the keys to their digital kingdom?

Why is Salesforce throwing humans at their human-replacing AI Agentforce? (make it make sense).

And does old school copyright mean a dang thing anymore? 

Don’t waste hours a day fretting over what the day’s AI news means. We do that for you each Monday with our AI News that Matters. 

Now, let's dive into this week's AI chaos and try to make sense of it, shall we? 

Here’s what ya need to know shorties.

The federal government just went full send on ChatGPT Enterprise.

According to reports from Fedscoop, The federal government is increasingly investing resources into the Enterprise version of ChatGPT.

NASA's locked in an annual license. The IRS dropped cash on 150 Enterprise seats for Treasury.

Los Alamos National Laboratory's joined the party too.

Even the Air Force Research Laboratory's getting cozy with OpenAI - their first Defense Department collab.

U.S. Agency for International Development was the OG federal customer.

What it means: 

The feds trusting OpenAI is like getting a five-star security rating from the Pentagon. If government agencies are cool with ChatGPT Enterprise, your company's "sensitive data" excuse just got weaker than WiFi at a tech conference.

Sid note — if your company is new to ChatGPT Enterprise, that’s something we help companies onboard. 

Salesforce is dropping major coin on human capital - 1,000+ new employees to sell AgentForce.

Their new AI agent handles customer support and sales development without human help.

CEO Mark Benoff's already flexing "incredible feedback" after two weeks.

Plot twist: Hiring humans to sell AI that replaces humans is such a human thing to do. 

Imagine the awkward onboarding. 

What it means: 

Salesforce just proved AI can't sell itself - yet. They're playing 4D chess: building AI to automate jobs while creating new ones to sell that automation.

Weird, right? 

But smart…. Yes? 

This move shows the real state of AI agents - powerful enough to disrupt in the long-run, but maybe not yet ready in the short-run. 

The Southern District of New York just handed OpenAI a big win, tossing out Raw Story Media and AlterNet Media’s lawsuit against them.

The judge's take? The plaintiffs couldn't prove OpenAI's training methods actually hurt them.

The court made it clear - ChatGPT creates new content rather than copying stuff word for word.

This ruling shows our copyright laws aren't quite ready for what AI's bringing to the table.

What it means: 

This is huge for companies training AI on web data. Without solid proof of copying or real damage, these lawsuits might keep hitting dead ends.

Content creators might need to switch strategies - think licensing deals instead of lawyer fees. The whole game's changing for how AI companies get their training data. 

(More on that in a bit) 

Donald Trump's second term as president could rewrite the AI rulebook from scratch, according to new reports. 

First thing likely to go? 

Biden's AI executive order, with the USAI Safety Institute looking shakier than a startup's revenue projections.

And what could add a wrench into Big Tech and AI companies alike? Trump’s proposed new tariffs could jack up AI hardware costs by 10-20% for imports and up to 60% for stuff coming from China.

Adding to the mix, Elon Musk might land a cabinet spot, bringing his AI doomsday vibes to Washington.

Yay?

What it means: 

Get ready for the AI policy pendulum to swing hard. Trump's crew is split between "full speed ahead" and "pump the brakes” so there’s no telling where the new administration will ultimately land on AI policy.

But it’s safe to say AI will be a focus, one way or another. 

Those import tariffs could make AI development cost more than a Tesla fleet. 

Microsoft's testing a new move - rolling out the features of Copilot Pro into Microsoft 365 subscriptions, but only in certain countries for now.

They're trying it out in Australia, New Zealand, Malaysia, Singapore, Taiwan, and Thailand first.

Instead of unlimited access like users would get with the $20/month Copilot Pro, users get monthly AI credits to play with.

There's a catch though - Microsoft 365 prices are going up to cover these new features.

In short, you pay a few more bucks a month instead of $20, but the limits are a bit lower. 

What it means: 

Looks like that separate $20/month for Copilot Pro wasn't hitting right? This bundled approach might be the future - AI features coming standard with your regular software subscription.

Watch these test markets closely - if users bite, this could be coming to your neck of the woods soon.

Right now, there’s no word on if this Copilot Pro remix strategy could hit the U.S. or European markets. 

Google accidentally dropped their Jarvis AI on the Chrome store before quickly yanking it back.

It's basically an AI sidekick that surfs the web with you.

According to reports, the official launch is set for December, but this leak gave us an early peek.

That name might need a rebrand though - just ask Jasper.ai about tangling with Marvel's lawyers. Lolz. 

What it means: 

Google's further along in the AI agent race than anyone thought. This leak shows they're not just playing around.

The Chrome integration could be their secret weapon - they've already got the browser market on lock.

Could this be a much more polished version of Anthropic’s Computer Use? 

We’d love to see it. 

Microsoft just dropped Magnetic One - a multi-agent system built on their Autogen framework.

It comes with four specialized AI agents: web surfer, file surfer, coder, and terminal wizard.

Right now, it runs on GPT-4 but plays nice with other AI models too. (We’re guessing it’ll get the o1 reasoning model from OpenAI soon.) 

Microsoft's pretty clear about one thing - this needs to be handled with care.

What it means: 

Microsoft's betting big on multi-agent systems being the next big thing. They're trying to get developers hooked on their framework before anyone else can.

Those safety warnings? Take em seriously - these agents pack more punch than your average AI tool.

OpenAI's Orion/GPT-5 model is hitting some interesting speedbumps in development, according to reporting from The Information. 

The next-gen model matches GPT-4's abilities after only 20% of training time - super efficient, but not the quantum leap forward we've gotten used to.

The company's not sleeping on this challenge though. They've pulled together a foundations team to solve their biggest headache: the shrinking supply of high-quality training data.

Their solution is straight out of left field. Instead of hunting for more real-world data, they're planning to feed Orion with synthetic data cooked up by their o1 reasoning models.

What it means: 

TBH, we’re not super convinced by this report. (And usually The Information is pretty spot on.)